Augmented Rotation-Based Transformation for Privacy-Preserving Data Clustering

摘要

Multiple rotation-based transformation (MRBT) was introduced recently formitigating the apriori-knowledge independent component analysis (AK-ICA) attackon rotation-based transformation (RBT), which is used for privacy-preservingdata clustering. MRBT is shown to mitigate the AK-ICA attack but at the expenseof data utility by not enabling conventional clustering. In this paper, weextend the MRBT scheme and introduce an augmented rotation-based transformation(ARBT) scheme that utilizes linearity of transformation and that both mitigatesthe AK-ICA attack and enables conventional clustering on data subsetstransformed using the MRBT. In order to demonstrate the computationalfeasibility aspect of ARBT along with RBT and MRBT, we develop a toolkit anduse it to empirically compare the different schemes of privacy-preserving dataclustering based on data transformation in terms of their overhead and privacy.